Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. An audio encoding apparatus for encoding a sequence of segments of coefficients, the segments being subsequent to each other in time, the audio encoding apparatus comprising a provider for providing the sequence of segments of coefficients from an audio stream representing a sampled audio signal by using different transform lengths such that segments of coefficients for which different transform lengths are used, spectrally represent the sampled audio signal at different frequency resolutions and comprise different numbers of coefficients; a processor for deriving an entropy coding context for a currently encoded coefficient of a current segment based on a previously encoded coefficient of a previous segment; and an entropy encoder for entropy encoding the current coefficient based on the entropy coding context to acquire an encoded audio stream, wherein the processor is configured to compute the entropy coding context for the current coefficient by selecting a set of coefficients of the previous segment in a manner so that in case the number of coefficients of the previous segment and the number of coefficients of the current segment are different, a number of coefficients in the set of coefficients is a first number, and in case the number of coefficients of the previous segment and the number of coefficients of the current segment are not different, the number of coefficients in the set of coefficients is a second number which is equal to the first number, and selecting at least some of the coefficients of the previous segment so that a spectral spacing between the selected coefficients of the previous segment is larger in case of the number of coefficients of the previous segment is larger as compared to a case that the number of coefficients of the previous segment the number of coefficients of the previous segment is smaller than the number of coefficients of the current segment, and computing the entropy coding context for the current coefficient on the basis of the set of coefficients.
This invention relates to audio encoding, specifically for encoding sequences of coefficient segments derived from an audio signal using variable transform lengths. The problem addressed is efficiently encoding audio data with different frequency resolutions, where segments of coefficients may have varying lengths and spectral representations. The apparatus includes a provider that generates segments of coefficients from an audio stream using different transform lengths, resulting in segments with different numbers of coefficients and varying frequency resolutions. A processor derives an entropy coding context for a current coefficient in the current segment based on a previously encoded coefficient from a previous segment. The processor selects a set of coefficients from the previous segment to compute the entropy coding context, adjusting the number of selected coefficients and their spectral spacing based on the relationship between the segment lengths. If the previous and current segments have different numbers of coefficients, the processor selects a fixed number of coefficients with adjusted spectral spacing to maintain consistency in context derivation. If the segment lengths are the same, the processor selects a different fixed number of coefficients. The entropy encoder then encodes the current coefficient using the derived context, producing an encoded audio stream. This approach ensures efficient entropy coding by adapting the context selection to varying segment lengths and frequency resolutions.
2. The audio encoding apparatus of claim 1 , wherein the entropy encoder is adapted for encoding the current coefficient in units of a tuple of spectral coefficients and for predicting a range of the tuple based on the entropy coding context.
This invention relates to audio encoding, specifically improving entropy encoding efficiency for spectral coefficients in audio signals. The problem addressed is the inefficiency in encoding spectral coefficients individually, which can lead to redundant data and increased bitrate. The solution involves encoding spectral coefficients in groups (tuples) rather than individually, while dynamically predicting the range of each tuple based on the entropy coding context. This context-aware prediction helps optimize the encoding process by reducing redundancy and improving compression efficiency. The apparatus includes an entropy encoder that processes these tuples, adjusting the encoding parameters based on the predicted range to enhance compression performance. By leveraging the relationships between adjacent spectral coefficients within a tuple, the method achieves more efficient encoding compared to traditional single-coefficient approaches. The invention is particularly useful in applications requiring high-quality audio compression, such as streaming and storage systems. The dynamic prediction of tuple ranges ensures adaptability to varying audio characteristics, further improving encoding efficiency.
3. The audio encoding apparatus of claim 2 , wherein the entropy encoder is adapted for dividing the tuple by a predetermined factor as often as necessitated to fit a result of the division in a predetermined range and for encoding a number of divisions necessitated, a division remainder and the result of the division when the tuple does not lie in the predicted range, and for encoding a division remainder and the result of the division otherwise.
This invention relates to audio encoding, specifically improving entropy encoding efficiency for audio data. The problem addressed is the inefficient encoding of large numerical values, such as tuples representing audio parameters, which can lead to excessive bit usage and reduced compression performance. The apparatus includes an entropy encoder that processes tuples of numerical values by dividing them by a predetermined factor when they fall outside a predicted range. The division is repeated until the result fits within a predefined range. The encoder then encodes the number of divisions performed, the remainder of the final division, and the resulting value. If the tuple is already within the predicted range, only the remainder and the result are encoded. This approach reduces bit usage by normalizing large values while preserving precision through the remainder and division count. The invention optimizes entropy encoding by dynamically adjusting the representation of numerical values based on their magnitude, improving compression efficiency without sacrificing accuracy. The method is particularly useful in audio codecs where precise yet compact representation of audio parameters is critical. The predetermined factor and range ensure consistent encoding behavior, while the division process adapts to the input data.
4. The audio encoding apparatus of claim 3 , wherein the entropy encoder is adapted for encoding the result of the division or the tuple using a group index, the group index referring to a group of one or more codewords for which a probability distribution is based on the entropy coding context, and, based on a uniform probability distribution, an element index in case the group comprises more than one codeword, the element index referring to a codeword within the group, and for encoding the number of divisions by a number of escape symbols, an escape symbol being a specific group index only used for indicating a division, and for encoding the remainders of the divisions based on a uniform probability distribution using an arithmetic coding rule.
This invention relates to audio encoding, specifically improving entropy encoding efficiency for audio data. The problem addressed is the inefficient representation of division results and tuples in entropy coding, which can lead to suboptimal compression performance. The solution involves an audio encoding apparatus that processes audio data by dividing it into segments and encoding the results using a structured approach. The apparatus includes an entropy encoder that encodes division results or tuples using a group index. The group index refers to a group of one or more codewords, where the probability distribution of the codewords is determined by the entropy coding context. If the group contains multiple codewords, an element index is used to specify the particular codeword within the group, encoded under a uniform probability distribution. The number of divisions is encoded using escape symbols, which are specific group indices reserved exclusively for indicating divisions. Remainders from the divisions are encoded using a uniform probability distribution under an arithmetic coding rule. This method ensures efficient and compact representation of encoded audio data, improving compression efficiency while maintaining accuracy.
5. The audio encoding apparatus of claim 4 , wherein the entropy encoder is adapted for encoding a sequence of symbols into the encoded audio stream using a symbol alphabet comprising the escape symbol, and group symbols corresponding to a set of available group indices, a symbol alphabet comprising the corresponding element indices, and a symbol alphabet comprising the different values of the remainders.
This invention relates to audio encoding, specifically improving efficiency in entropy encoding of audio data. The problem addressed is the need for more compact representation of audio parameters, particularly when encoding sequences of symbols that include repeated or predictable patterns. The apparatus includes an entropy encoder that processes a sequence of symbols into an encoded audio stream. The encoder uses multiple symbol alphabets to optimize encoding. One alphabet includes an escape symbol and group symbols corresponding to available group indices, allowing efficient representation of frequently occurring groups. Another alphabet contains element indices, enabling precise encoding of individual elements within groups. A third alphabet encodes different values of remainders, handling cases where values do not fit neatly into predefined groups. The encoder dynamically selects symbols from these alphabets based on the input data, reducing redundancy and improving compression efficiency. The system is particularly useful in audio codecs where parameter sequences exhibit statistical patterns that can be exploited for better compression. The invention ensures that both common and rare values are encoded efficiently, balancing between group-based and individual encoding to minimize bitrate while maintaining accuracy.
6. The audio encoding apparatus of claim 1 , wherein the processor and the entropy encoder are configured to operate based on a down-sampling of spectral coefficients of the previous segment, when the previous segment displays a finer spectral resolution than the current segment and/or wherein the processor and the entropy encoder are configured to operate based on an up-sampling of spectral coefficients of the previous segment, when the previous segment displays a coarser spectral resolution than the current segment.
This invention relates to audio encoding, specifically improving efficiency in entropy encoding by dynamically adjusting spectral resolution between consecutive audio segments. The problem addressed is the computational and bandwidth overhead when encoding audio segments with varying spectral resolutions, such as in adaptive transform coding or hybrid coding schemes. The apparatus includes a processor and an entropy encoder that handle spectral coefficients of audio segments. When the previous segment has a finer spectral resolution than the current segment, the processor and encoder down-sample the spectral coefficients of the previous segment to match the current segment's resolution before encoding. Conversely, if the previous segment has a coarser resolution, the processor and encoder up-sample the spectral coefficients to align with the current segment's resolution. This adaptive resolution adjustment ensures consistent encoding efficiency by preventing mismatches in spectral granularity between adjacent segments, reducing redundancy and improving compression performance. The method applies to any audio encoding system where spectral resolution may vary dynamically, such as in perceptual audio codecs or speech coding.
7. The audio encoding apparatus of claim 1 , wherein the number of coefficients in the set of coefficients is four.
This invention relates to audio encoding, specifically improving efficiency in digital audio compression. The problem addressed is the computational overhead and memory usage in encoding audio signals, particularly when processing large sets of coefficients representing audio data. Traditional methods often require extensive processing to encode these coefficients, leading to inefficiencies in both time and resources. The invention describes an audio encoding apparatus that includes a coefficient selector and an encoder. The coefficient selector extracts a set of coefficients from an audio signal, where the number of coefficients in the set is fixed at four. This fixed-size approach simplifies the encoding process by reducing the variability in data handling. The encoder then processes these four coefficients to generate an encoded output, optimizing compression while maintaining audio quality. By limiting the coefficient set to four, the apparatus ensures consistent processing requirements, minimizing computational complexity and memory usage. This design is particularly useful in real-time audio applications where efficiency is critical, such as streaming, voice communication, or embedded audio systems. The fixed-size coefficient set allows for predictable performance, making it easier to implement in hardware or software with constrained resources. The invention thus provides a balanced solution between compression efficiency and processing simplicity in audio encoding.
8. The audio encoding apparatus according to claim 1 wherein the entropy encoder uses the entropy coding context to arithmetically encode the current coefficient.
This invention relates to audio encoding, specifically improving the efficiency of entropy encoding in audio compression systems. The problem addressed is the need to reduce bitrate while maintaining audio quality by optimizing the way audio coefficients are encoded. The apparatus includes a coefficient analyzer that determines the significance of an audio coefficient in a frequency domain representation, such as a transform domain like MDCT (Modified Discrete Cosine Transform). The analyzer checks whether the coefficient is significant, meaning it exceeds a predefined threshold, and if so, determines its sign and magnitude. The entropy encoder then uses an entropy coding context, which is derived from neighboring coefficients or other statistical properties, to arithmetically encode the current coefficient. The context selection ensures that the encoding process adapts to local signal characteristics, improving compression efficiency. The apparatus may also include a bit allocation module that assigns bits based on perceptual importance, further optimizing the encoding process. The overall system aims to minimize bitrate while preserving audio fidelity by leveraging adaptive entropy coding and perceptual modeling.
9. The audio encoding apparatus according to claim 8 , wherein the previous segment and current segment are spectrally subdivided into tuples of spectrally neighbouring coefficients, respectively, with a number of coefficients per tuple being equal for tuples of the previous segments and tuples for the current segment, wherein the processor is configured to compute the entropy coding context for the current coefficient by selecting a set of tuples of the previous segment in a manner so that a number of coefficients in the set of tuples in case of the number of coefficients of the previous segment and the number of coefficients of the current segment being different, is equal to the number of coefficients in the set of coefficients in in case of the number of coefficients of the previous segment and the number of coefficients of the current segment not being different, and at least some of the set of tuples are selected out of the tuples of the previous segment and selected in a manner so that a spectral spacing between coefficients of the set of coefficients selected out of the coefficients of the previous segment is larger in case of the number of coefficients of the previous segment and being larger than the number of coefficients of the current segment than in case of the number of coefficients of the previous segment being lower than the number of coefficients of the current segment, and computing the entropy coding context for the current coefficient on the basis of the number of coefficients of each of the set of tuples, and the entropy encoder is configured to entropy encode the current coefficient by entropy encoding a tuple which comprises the current coefficient using the entropy coding context.
Audio encoding systems process audio signals by converting them into a compressed format for efficient storage or transmission. A challenge in such systems is accurately predicting and encoding spectral coefficients, especially when the number of coefficients varies between segments of the audio signal. This can lead to inefficiencies in entropy coding, where context selection becomes less accurate, degrading compression performance. An audio encoding apparatus addresses this issue by spectrally subdividing both a previous segment and a current segment of the audio signal into tuples of spectrally neighboring coefficients. The number of coefficients per tuple is consistent across both segments. The apparatus computes an entropy coding context for a current coefficient by selecting a set of tuples from the previous segment. If the number of coefficients in the previous and current segments differs, the selection ensures the total number of coefficients in the set remains equal to the case where the segment sizes match. Additionally, the selection prioritizes tuples with larger spectral spacing when the previous segment has more coefficients than the current segment, and smaller spacing when the opposite is true. The entropy coding context is then derived from the number of coefficients in each selected tuple. The entropy encoder uses this context to encode the current coefficient as part of its tuple, improving compression efficiency by adapting to varying segment sizes. This method enhances spectral coefficient prediction and encoding accuracy, particularly in scenarios where segment sizes fluctuate.
10. The audio encoding apparatus of claim 9 , wherein the number of coefficients in the set of coefficients is four.
This invention relates to audio encoding, specifically improving efficiency in transforming audio signals into a compressed format. The problem addressed is the computational complexity and memory usage in audio encoding, particularly when processing audio signals using transform-based methods like the Modified Discrete Cosine Transform (MDCT). Traditional approaches often require large sets of coefficients, leading to inefficiencies in processing and storage. The invention describes an audio encoding apparatus that includes a coefficient selection unit and a transformation unit. The coefficient selection unit selects a set of coefficients from a predefined set, where the number of coefficients in the selected set is four. The transformation unit then applies a transformation to an audio signal using the selected set of coefficients to generate a transformed audio signal. This transformation is designed to reduce computational overhead and memory requirements by limiting the number of coefficients used, while still maintaining audio quality. The apparatus may also include a quantization unit that quantizes the transformed audio signal and an entropy encoding unit that encodes the quantized signal. The coefficient selection unit can dynamically adjust the selected coefficients based on the characteristics of the input audio signal, such as frequency content or energy distribution, to optimize encoding efficiency. The use of a fixed number of coefficients (four) simplifies the implementation and ensures consistent performance across different audio signals. This approach is particularly useful in real-time audio encoding applications where processing speed and resource efficiency are critical.
11. Method for encoding a sequence of segments of coefficients, the segments being subsequent to each other in time, the method comprising providing the sequence of segments of coefficients from an audio stream representing a sampled audio signal by using different transform lengths such that segments of coefficients for which different transform lengths are used, spectrally represent the sampled audio signal at different frequency resolutions and comprise different numbers of coefficients; deriving an entropy coding context for a currently encoded coefficient of a current segment based on a previously encoded coefficient of a previous segment; and entropy encoding the current coefficient based on the entropy coding context to acquire an encoded audio stream, wherein the deriving the entropy coding context comprises computing the entropy coding context for the current coefficient by selecting a set of coefficients of the previous segment in a manner so that in case the number of coefficients of the previous segment and the number of coefficients of the current segment are different, a number of coefficients in the set of coefficients is a first number, and in case the number of coefficients of the previous segment and the number of coefficients of the current segment are not different, the number of coefficients in the set of coefficients is a second number which is equal to the first number, and selecting at least some of the coefficients of the previous segment so that a spectral spacing between the selected coefficients of the previous segment is larger in case of the number of coefficients of the previous segment is larger as compared to a case that the number of coefficients of the previous segment the number of coefficients of the previous segment is smaller than the number of coefficients of the current segment, and computing the entropy coding context for the current coefficient on the basis of the set of coefficients.
This method relates to audio signal encoding, specifically for efficiently compressing audio streams using variable transform lengths. The technique addresses the challenge of maintaining high compression efficiency while adapting to different frequency resolutions in the audio signal. The method processes a sequence of coefficient segments derived from an audio stream, where each segment is obtained using different transform lengths, resulting in varying numbers of coefficients and spectral resolutions. To improve entropy coding efficiency, the method derives an entropy coding context for a current coefficient in the current segment based on previously encoded coefficients from a prior segment. The context derivation involves selecting a subset of coefficients from the previous segment, where the selection adapts to the difference in the number of coefficients between the current and previous segments. If the segments have different numbers of coefficients, a fixed number of coefficients are selected, while if they have the same number, the same fixed number is still used. Additionally, the selection ensures that the spectral spacing between the chosen coefficients is larger when the previous segment has more coefficients than the current segment, and smaller when it has fewer. The entropy coding context is then computed based on this selected subset, and the current coefficient is encoded accordingly, producing an efficiently compressed audio stream. This approach optimizes entropy coding by dynamically adjusting context selection based on varying transform lengths, improving compression performance for audio signals with time-varying frequency characteristics.
12. The method according to claim 11 wherein the entropy encoding comprises arithmetically encoding the current coefficient.
This invention relates to video encoding, specifically improving compression efficiency by optimizing entropy encoding of transform coefficients. In video compression, transform coefficients are generated after applying a transform (e.g., DCT) to residual blocks, and these coefficients are then entropy-encoded to reduce bitrate. A challenge in entropy encoding is efficiently representing coefficients with varying statistical properties, particularly in high-frequency regions where coefficients may be sparse or have low magnitude. The method addresses this by using arithmetic encoding for the current coefficient, which provides better compression than simpler methods like Huffman coding, especially for coefficients with non-uniform distributions. Arithmetic encoding assigns variable-length codes based on probability models, allowing finer granularity in bit allocation. The method may involve adaptive probability models that update based on previously encoded coefficients or context information, further improving efficiency. Additionally, the method may include coefficient scanning techniques to group similar coefficients before encoding, enhancing the effectiveness of arithmetic encoding. By combining these techniques, the method achieves higher compression efficiency while maintaining or improving reconstruction quality.
13. The method for encoding a sequence of segments of coefficients of claim 11 , wherein the number of coefficients in the set of coefficients is four.
This invention relates to a method for encoding a sequence of segments of coefficients, specifically where each segment contains exactly four coefficients. The method is designed for efficient data compression, particularly in applications involving coefficient-based representations such as image or signal processing. The problem addressed is the need for optimized encoding of coefficient segments to reduce storage or transmission overhead while maintaining data integrity. The method involves processing a sequence of segments, where each segment consists of four coefficients. These coefficients may represent transformed data, such as discrete cosine transform (DCT) coefficients in image compression. The encoding process likely includes steps to analyze the coefficients, identify patterns, and apply a compression scheme tailored to the segment structure. This may involve techniques like entropy coding, quantization, or other lossless or lossy compression methods to minimize redundancy. The invention ensures that the encoding process is efficient and scalable, particularly when dealing with fixed-size segments of four coefficients. This approach may be used in video coding, audio compression, or other domains where coefficient-based representations are common. The method aims to balance compression efficiency with computational complexity, making it suitable for real-time or resource-constrained environments. The fixed segment size of four coefficients simplifies the encoding logic and may improve hardware or software implementation efficiency.
14. A non-transitory computer-readable storage medium storing a computer program comprising a program code for performing the method according to claim 11 when the program code runs on a computer or a processor.
The invention relates to a computer program stored on a non-transitory computer-readable storage medium, designed to execute a method for optimizing data processing in a distributed computing environment. The method involves analyzing data distribution across multiple nodes to identify imbalances, then dynamically redistributing the data to balance the computational load. This ensures efficient resource utilization and reduces processing delays. The program code, when executed, monitors data flow, detects bottlenecks, and adjusts data allocation in real-time to maintain optimal performance. The system is particularly useful in large-scale distributed systems where uneven data distribution can lead to inefficiencies. The storage medium may include any non-volatile memory device capable of retaining the program code for execution on a computer or processor. The method ensures that data is evenly distributed, preventing overloading of specific nodes and improving overall system throughput. The invention addresses the challenge of maintaining balanced workloads in distributed computing environments, where traditional static allocation methods often fail to adapt to dynamic data changes. By dynamically redistributing data, the system enhances scalability and reliability, making it suitable for applications requiring high-performance computing.
15. The method according to claim 11 , wherein the previous segment and current segment are spectrally subdivided into tuples of spectrally neighbouring coefficients, respectively, with a number of coefficients per tuple being equal for tuples of the previous segments and tuples for the current segment, wherein the deriving the entropy coding context comprises computing the entropy coding context for the current coefficient by selecting a set of tuples of the previous segment in a manner so that a number of coefficients in the set of tuples in case of the number of coefficients of the previous segment and the number of coefficients of the current segment being different, is equal to the number of coefficients in the set of coefficients in case of the number of coefficients of the previous segment and the number of coefficients of the current segment not being different, and at least some of the set of tuples are selected out of the tuples of the previous segment and selected in a manner so that a spectral spacing between coefficients of the set of coefficients selected out of the coefficients of the previous segment is larger in case of the number of coefficients of the previous segment and being larger than the number of coefficients of the current segment than in case of the number of coefficients of the previous segment being lower than the number of coefficients of the current segment, and computing the entropy coding context for the current coefficient on the basis of the number of coefficients of each of the set of tuples, and the entropy encoding comprises entropy encoding the current coefficient by entropy encoding a tuple which comprises the current coefficient using the entropy coding context.
This invention relates to spectral data processing, specifically methods for entropy coding coefficients in audio or video compression. The problem addressed is efficiently encoding spectral coefficients when the number of coefficients in a current segment differs from a previous segment, ensuring consistent entropy coding context derivation. The method involves spectrally subdividing both the previous and current segments into tuples of neighboring coefficients, where each tuple contains an equal number of coefficients across both segments. To derive the entropy coding context for a current coefficient, a set of tuples from the previous segment is selected. If the previous and current segments have different numbers of coefficients, the selection ensures the total number of coefficients in the set matches the case where the segment sizes are equal. Additionally, the selection prioritizes tuples with larger spectral spacing between coefficients when the previous segment has more coefficients than the current segment, and smaller spacing when the opposite is true. The entropy coding context is then computed based on the number of coefficients in each selected tuple. Finally, the current coefficient is entropy-encoded by encoding its containing tuple using the derived context. This approach maintains coding efficiency and consistency despite varying segment sizes.
16. The method for encoding a sequence of segments of coefficients of claim 15 , wherein the number of coefficients in the set of coefficients is four.
This invention relates to a method for encoding a sequence of segments of coefficients, specifically where the number of coefficients in each set is four. The method addresses the challenge of efficiently encoding coefficient data, which is common in signal processing, image compression, and other data transformation applications. By segmenting the coefficients into groups of four, the encoding process can be optimized for both computational efficiency and data compression. The method involves processing each segment of four coefficients to generate encoded data, which can then be transmitted, stored, or further processed. The segmentation approach helps reduce redundancy and improves encoding accuracy, particularly in applications where coefficient values exhibit spatial or temporal correlations. The method may be applied in various domains, including but not limited to video encoding, audio compression, and machine learning model optimization. By standardizing the coefficient set size to four, the encoding process becomes more predictable and easier to implement in hardware or software systems. This approach ensures consistent performance across different data types and processing environments.
17. An audio decoding apparatus for decoding an encoded audio stream representing a sampled audio signal to acquire a sequence of segments of coefficients being subsequent to each other in time and representing the sampled audio signal by using different transform lengths such that segments of coefficients for which different transform lengths are used, spectrally represent the sampled audio signal at different frequency resolutions and comprise different numbers of coefficients, comprising a processor for deriving an entropy coding context for a currently decoded coefficient of a current segment based on a previously decoded coefficient of a previous segment; and an entropy decoder for entropy decoding the current coefficient based on the entropy coding context and the encoded audio stream, wherein the processor is configured to compute the entropy coding context for the current coefficient by selecting a set of coefficients of the previous segment in a manner so that in case the number of coefficients of the previous segment and the number of coefficients of the current segment are different, a number of coefficients in the set of coefficients is a first number, and in case the number of coefficients of the previous segment and the number of coefficients of the current segment are not different, the number of coefficients in the set of coefficients is a second number which is equal to the first number, and selecting at least some of the coefficients of the previous segment so that a spectral spacing between the selected coefficients of the previous segment is larger in case of the number of coefficients of the previous segment is larger as compared to a case that the number of coefficients of the previous segment the number of coefficients of the previous segment is smaller than the number of coefficients of the current segment, and computing the entropy coding context for the current coefficient on the basis of the set of coefficients.
This invention relates to audio decoding, specifically for handling encoded audio streams with variable transform lengths. The problem addressed is efficiently decoding audio segments that use different transform lengths, resulting in varying frequency resolutions and different numbers of coefficients per segment. The solution involves an audio decoding apparatus that processes an encoded audio stream to reconstruct a sampled audio signal by decoding segments of coefficients with adaptive entropy coding contexts. The apparatus includes a processor and an entropy decoder. The processor derives an entropy coding context for a currently decoded coefficient in a current segment by analyzing a previously decoded segment. The context is computed by selecting a set of coefficients from the previous segment, where the number of selected coefficients depends on whether the current and previous segments have the same number of coefficients. If the segment sizes differ, a fixed number of coefficients is chosen; if they match, the same fixed number is used. Additionally, the selection ensures that the spectral spacing between selected coefficients is larger when the previous segment has more coefficients, optimizing context derivation for varying resolutions. The entropy decoder then uses this context to decode the current coefficient from the encoded stream. This approach improves decoding efficiency by adapting to changes in transform length while maintaining consistent context-based entropy decoding.
18. The audio decoding apparatus of claim 17 , wherein the processor is adapted for deriving the entropy coding context per spectral band for the current coefficient, based on neighbouring spectral coefficients previously decoded in one or more of the previous segment and the present segment.
This invention relates to audio decoding, specifically improving entropy coding efficiency by optimizing context selection for spectral coefficients. The problem addressed is the inefficiency in traditional audio decoding where entropy coding contexts are selected without leveraging neighboring spectral information, leading to suboptimal compression performance. The apparatus includes a processor that decodes spectral coefficients from an encoded audio signal, where the audio signal is divided into segments. The processor derives an entropy coding context for a current spectral coefficient by analyzing neighboring spectral coefficients that have already been decoded. These neighboring coefficients may come from the same segment as the current coefficient or from a previous segment. By using this contextual information, the entropy coding process becomes more accurate, improving compression efficiency and reducing bitrate without sacrificing audio quality. The processor may also apply a spectral weighting function to the neighboring coefficients before deriving the context, ensuring that closer or more relevant coefficients have a greater influence on the context selection. This adaptive approach enhances the precision of entropy coding, particularly in complex audio signals where spectral characteristics vary significantly across segments. The method ensures that the context selection dynamically adapts to the audio signal's structure, optimizing decoding performance.
19. The audio decoding apparatus of claim 18 , wherein the entropy decoder is adapted for decoding a group index from the encoded audio stream based on a probability distribution derived from the entropy coding context, wherein the group index represents a group of one or more codewords, and for, based on a uniform probability distribution, decoding an element index from the encoded audio stream if the group index indicates a group comprising more than one codeword, and for deriving a tuple of spectral coefficients of the current segment based on the group index and the element index, thereby acquiring the spectral domain representation in tuples of spectral coefficients.
This invention relates to audio decoding, specifically improving the efficiency of entropy decoding in spectral domain audio processing. The problem addressed is the computational overhead and complexity in decoding spectral coefficients from an encoded audio stream, particularly when handling groups of codewords with varying probabilities. The audio decoding apparatus includes an entropy decoder that processes an encoded audio stream to reconstruct spectral coefficients. The entropy decoder first decodes a group index from the stream using a probability distribution derived from an entropy coding context. The group index identifies a group of one or more codewords. If the group contains multiple codewords, the decoder further decodes an element index using a uniform probability distribution. The group index and element index are then combined to derive a tuple of spectral coefficients for the current audio segment. This approach reduces decoding complexity by leveraging context-based probability distributions for group selection and uniform distributions for finer-grained codeword selection within groups, optimizing both speed and accuracy in spectral domain reconstruction. The method ensures efficient decoding while maintaining high-quality audio output.
20. The audio decoding apparatus of claim 19 , wherein the entropy decoder is adapted for decoding a sequence of symbols from the encoded audio stream based on the probability distribution derived from the entropy coding context using a symbol alphabet comprising an escape symbol and group symbols corresponding to a set of available group indices, for deriving a preliminary tuple of spectral coefficients based on an available group index to which a group symbol of the sequence of symbols corresponds, and based on the element index, and for multiplying the preliminary tuple with a factor depending on a number of escape symbols in the sequence of symbols to acquire the tuple of spectral coefficients.
Audio decoding systems process encoded audio streams to reconstruct spectral coefficients for audio synthesis. A key challenge is efficiently decoding symbols from the encoded stream while maintaining high compression efficiency and accurate spectral reconstruction. This invention addresses this by improving the entropy decoding process in audio decoding apparatuses. The apparatus includes an entropy decoder that decodes a sequence of symbols from the encoded audio stream using a probability distribution derived from an entropy coding context. The symbol alphabet includes an escape symbol and group symbols, where each group symbol corresponds to a set of available group indices. The decoder derives a preliminary tuple of spectral coefficients based on the group index associated with a group symbol in the sequence and an element index. The preliminary tuple is then scaled by a factor determined by the number of escape symbols in the sequence to produce the final tuple of spectral coefficients. This approach optimizes the decoding process by dynamically adjusting the spectral coefficients based on the presence of escape symbols, improving both efficiency and accuracy in reconstructing the audio signal.
21. The audio decoding apparatus of claim 20 , wherein the entropy decoder is adapted for decoding a division remainder from the encoded audio stream based on a uniform probability distribution using an arithmetic coding rule and for adding the remainder to the multiplied preliminary tuple to acquire the tuple of spectral coefficients.
The invention relates to audio decoding, specifically improving the efficiency of decoding spectral coefficients from an encoded audio stream. The problem addressed is the computational complexity and accuracy in reconstructing audio signals from encoded data, particularly when dealing with division operations that can introduce rounding errors or require significant processing power. The audio decoding apparatus includes an entropy decoder that processes an encoded audio stream to reconstruct spectral coefficients. The decoder first retrieves a preliminary tuple of spectral coefficients, which is then multiplied by a scaling factor. To refine this result, the entropy decoder decodes a division remainder from the encoded stream using an arithmetic coding rule that assumes a uniform probability distribution. This remainder is then added to the scaled preliminary tuple to produce the final tuple of spectral coefficients. This approach ensures precise reconstruction while minimizing computational overhead by avoiding direct division operations, which are computationally expensive. The method leverages arithmetic coding to efficiently encode and decode the remainder, improving both accuracy and processing efficiency in audio decoding.
22. The audio decoding apparatus of claim 21 , wherein the processor and the entropy encoder are configured to operate based on a down-sampling of spectral coefficients of the previous segment, when the previous segment displays a finer spectral resolution than the current segment and/or wherein the processor and the entropy encoder are configured to operate based on an up-sampling of spectral coefficients of the previous segment, when the previous segment displays a coarser spectral resolution than the current segment.
This invention relates to audio decoding systems that handle spectral coefficients with varying resolutions between adjacent audio segments. The problem addressed is the inefficiency in encoding and decoding audio signals when segments have different spectral resolutions, leading to increased computational complexity and potential artifacts. The apparatus includes a processor and an entropy encoder that adaptively adjust spectral coefficients between a current audio segment and a previous segment. When the previous segment has a finer spectral resolution than the current segment, the processor and encoder down-sample the spectral coefficients of the previous segment to match the resolution of the current segment. Conversely, if the previous segment has a coarser resolution, the processor and encoder up-sample the spectral coefficients to align with the current segment's resolution. This dynamic adjustment ensures smooth transitions between segments with different resolutions, improving encoding efficiency and audio quality. The system avoids redundant computations and reduces artifacts by ensuring consistent spectral resolution handling, particularly in scenarios where adaptive bitrate or variable resolution encoding is used. The processor and entropy encoder work together to maintain coherence in the decoded audio signal, optimizing both computational resources and perceptual fidelity.
23. The audio decoding apparatus according to claim 17 wherein the entropy decoder uses the entropy coding context to arithmetically decode the current coefficient.
This invention relates to audio decoding, specifically improving the efficiency of entropy decoding in audio codecs. The problem addressed is the computational overhead and inefficiency in decoding audio coefficients, particularly in systems where entropy coding is used to compress audio data. The invention provides an audio decoding apparatus that includes an entropy decoder configured to use an entropy coding context to arithmetically decode a current coefficient. The entropy coding context is derived from previously decoded coefficients or other contextual information, allowing the decoder to select an optimal arithmetic decoding method for the current coefficient. This reduces redundancy and improves decoding speed and accuracy. The apparatus may also include a coefficient reconstruction module that processes the decoded coefficients to reconstruct the audio signal. The entropy coding context may be updated dynamically based on the decoded coefficients to adapt to changes in the audio signal. This adaptive approach ensures efficient decoding across different audio segments, enhancing overall performance. The invention is particularly useful in real-time audio applications where low-latency decoding is critical.
24. The audio decoding apparatus of claim 17 , wherein the number of coefficients in the set of coefficients is four.
This invention relates to audio decoding, specifically improving efficiency in processing audio signals. The problem addressed is the computational overhead in audio decoding, particularly when handling large sets of coefficients used in signal reconstruction. Traditional methods require extensive processing to decode audio data, leading to delays and increased power consumption, which is problematic for real-time applications and portable devices. The invention provides an audio decoding apparatus that optimizes the decoding process by using a set of coefficients to reconstruct an audio signal. The apparatus includes a coefficient storage unit that holds a predefined set of coefficients, a coefficient selection unit that selects a subset of these coefficients based on the input audio data, and a signal reconstruction unit that applies the selected coefficients to reconstruct the audio signal. The apparatus further includes a control unit that manages the selection and application of coefficients to ensure accurate and efficient decoding. A key feature of this invention is that the number of coefficients in the set is limited to four. This reduction in the number of coefficients minimizes computational complexity while maintaining audio quality, making the decoding process faster and more energy-efficient. The apparatus is particularly useful in applications where low latency and power efficiency are critical, such as mobile devices, real-time audio streaming, and voice recognition systems. By restricting the coefficient set to four, the invention balances performance and resource usage, addressing the need for efficient audio decoding in modern digital systems.
25. The audio decoding apparatus according to claim 17 , wherein the previous segment and current segment are spectrally subdivided into tuples of spectrally neighbouring coefficients, respectively, with a number of coefficients per tuple being equal for tuples of the previous segments and tuples for the current segment, wherein the processor is configured to compute the entropy coding context for the current coefficient by selecting a set of tuples of the previous segment in a manner so that a number of coefficients in the set of tuples in case of the number of coefficients of the previous segment and the number of coefficients of the current segment being different, is equal to the number of coefficients in the set of coefficients in case of the number of coefficients of the previous segment and the number of coefficients of the current segment not being different, and at least some of the set of tuples are selected out of the tuples of the previous segment and selected in a manner so that a spectral spacing between coefficients of the set of coefficients selected out of the coefficients of the previous segment is larger in case of the number of coefficients of the previous segment and being larger than the number of coefficients of the current segment than in case of the number of coefficients of the previous segment being lower than the number of coefficients of the current segment, and computing the entropy coding context for the current coefficient on the basis of the number of coefficients of each of the set of tuples, and the entropy decoder is configured to entropy decode the current coefficient by entropy decoding a tuple which comprises the current coefficient using the entropy coding context.
Audio decoding involves processing spectrally subdivided audio segments to reconstruct the original signal efficiently. A key challenge is accurately encoding and decoding coefficients in segments with varying lengths while maintaining spectral coherence. This invention addresses this by improving entropy coding context selection for audio coefficients. The apparatus processes audio segments divided into tuples of spectrally neighboring coefficients, ensuring each tuple has the same number of coefficients across segments. When encoding a current coefficient, the processor selects a set of tuples from a previous segment. If the previous and current segments have different numbers of coefficients, the selection ensures the set contains the same number of coefficients as it would if the segments had equal lengths. Additionally, the selection prioritizes tuples with larger spectral spacing when the previous segment has more coefficients than the current segment, and smaller spacing when it has fewer. The entropy coding context for the current coefficient is then computed based on the number of coefficients in each selected tuple. The entropy decoder uses this context to decode the current coefficient within its tuple. This method ensures consistent and efficient entropy coding across segments of varying lengths, improving decoding accuracy and spectral fidelity.
26. The audio decoding apparatus of claim 25 , wherein the number of coefficients in the set of coefficients is four.
The invention relates to audio decoding, specifically improving the efficiency and accuracy of audio signal reconstruction. The problem addressed is the computational complexity and potential quality loss in traditional audio decoding methods, particularly when processing coefficients used in signal reconstruction. The apparatus includes a coefficient processor that selects a set of coefficients from an encoded audio signal, where the number of coefficients in the set is four. These coefficients are used to reconstruct the audio signal with improved accuracy and reduced computational overhead. The coefficient processor may also apply a transformation or filtering operation to the selected coefficients to enhance the reconstruction process. The apparatus further includes a signal generator that uses the processed coefficients to generate the decoded audio signal, ensuring high fidelity while minimizing processing resources. This approach optimizes the balance between computational efficiency and audio quality, making it suitable for real-time applications such as streaming and portable audio devices.
27. A method for decoding an encoded audio stream representing a sampled audio signal to acquire a sequence of segments of coefficients being subsequent to each other in time and representing the sampled audio signal by using different transform lengths such that segments of coefficients for which different transform lengths are used, spectrally represent the sampled audio signal at different frequency resolutions and comprise different numbers of coefficients, comprising deriving an entropy coding context for a currently decoded coefficient of a current segment based on a previously decoded coefficient of a previous segment; and entropy decoding the current coefficient based on the entropy coding context and the encoded audio stream, wherein the deriving the entropy coding context comprises computing the entropy coding context for the current coefficient by selecting a set of coefficients of the previous segment in a manner so that in case the number of coefficients of the previous segment and the number of coefficients of the current segment are different, a number of coefficients in the set of coefficients is a first number, and in case the number of coefficients of the previous segment and the number of coefficients of the current segment are not different, the number of coefficients in the set of coefficients is a second number which is equal to the first number, and selecting at least some of the coefficients of the previous segment so that a spectral spacing between the selected coefficients of the previous segment is larger in case of the number of coefficients of the previous segment is larger as compared to a case that the number of coefficients of the previous segment the number of coefficients of the previous segment is smaller than the number of coefficients of the current segment, and computing the entropy coding context for the current coefficient on the basis of the set of coefficients.
This invention relates to audio decoding, specifically methods for decoding an encoded audio stream that uses variable transform lengths to represent different segments of the audio signal at varying frequency resolutions. The problem addressed is efficiently decoding coefficients in segments with different lengths while maintaining accurate entropy coding contexts for optimal compression and reconstruction. The method involves decoding an encoded audio stream into a sequence of coefficient segments, where each segment may have a different number of coefficients due to varying transform lengths. The key innovation is in deriving an entropy coding context for a current coefficient in the current segment based on previously decoded coefficients from a prior segment. The context derivation adapts to the differing segment lengths by selecting a subset of coefficients from the previous segment. If the previous and current segments have different numbers of coefficients, a fixed number of coefficients are selected from the previous segment, ensuring consistency in context derivation. If the segment lengths are the same, the same fixed number is used. Additionally, the selection of coefficients from the previous segment adjusts the spectral spacing between them based on the relative sizes of the segments, ensuring that larger segments have wider spacing between selected coefficients. The entropy coding context for the current coefficient is then computed using this selected subset, enabling efficient entropy decoding of the current coefficient. This approach ensures robust decoding even when segment lengths vary, improving compression efficiency and audio quality.
28. The method according to claim 27 wherein the entropy decoding comprises arithmetically decoding the current coefficient.
This invention relates to video encoding and decoding, specifically improving entropy decoding efficiency in video compression. The problem addressed is the computational overhead in decoding quantized transform coefficients, particularly in arithmetic decoding, which can slow down real-time video processing. The method involves decoding a current coefficient in a video frame using arithmetic decoding. The current coefficient is part of a block of quantized transform coefficients, which are derived from applying a transform (e.g., DCT) to residual pixel data. The arithmetic decoding process involves entropy decoding, where symbols representing the coefficients are decoded using an arithmetic coding scheme, which is more efficient than fixed-length or variable-length coding for high-compression scenarios. Before arithmetic decoding, the method may include reconstructing a context model for the current coefficient based on previously decoded coefficients or neighboring blocks. The context model helps predict the probability distribution of the current coefficient, improving decoding accuracy. The arithmetic decoder then uses this model to decode the current coefficient, reducing the number of bits required compared to traditional methods. The method may also involve inverse quantization and inverse transformation after decoding, where the quantized coefficients are scaled back to their original range and transformed back into the spatial domain to reconstruct the residual block. This residual is then added to a predicted block to form the final reconstructed video frame. The invention aims to optimize entropy decoding in video compression, particularly in high-efficiency video coding (HEVC) or similar standards, by leveraging arithmetic decoding for efficient coeffici
29. A non-transitory computer-readable storage medium storing a computer program comprising a program code for performing the method according to claim 27 when the program code runs on a computer or a processor.
The invention relates to a computer program stored on a non-transitory computer-readable storage medium, designed to execute a method for optimizing data processing in a distributed computing environment. The method involves analyzing data distribution across multiple nodes to identify imbalances, dynamically redistributing data to balance the load, and adjusting processing tasks based on the redistributed data. The program code, when executed, enables a computer or processor to perform these steps, ensuring efficient resource utilization and minimizing processing delays. The method includes monitoring data access patterns, determining optimal data placement strategies, and reallocating data to nodes with lower processing loads. This approach improves system performance by reducing bottlenecks and enhancing parallel processing capabilities. The invention addresses the challenge of uneven data distribution in distributed systems, which can lead to inefficient resource usage and slower processing times. By dynamically balancing data across nodes, the system achieves more consistent performance and better scalability. The computer program is designed to be executed on standard computing hardware, making it adaptable to various distributed computing environments.
30. The method for decoding an encoded audio stream of claim 27 , wherein the number of coefficients in the set of coefficients is four.
The invention relates to audio signal processing, specifically methods for decoding an encoded audio stream. The problem addressed is improving the efficiency and accuracy of audio decoding by optimizing the number of coefficients used in the decoding process. The method involves decoding an encoded audio stream by processing a set of coefficients derived from the encoded data. The key innovation is specifying that the set of coefficients used in the decoding process consists of exactly four coefficients. This fixed number of coefficients ensures a balance between computational efficiency and audio quality, reducing the complexity of the decoding algorithm while maintaining high-fidelity audio reconstruction. The method may be applied in various audio decoding systems, including real-time streaming and playback devices, where efficient processing is critical. By limiting the coefficient set to four, the method avoids unnecessary computations and memory usage, making it suitable for resource-constrained environments. The approach is particularly useful in scenarios where audio quality must be preserved without excessive processing overhead.
31. The method according to claim 27 , wherein the previous segment and current segment are spectrally subdivided into tuples of spectrally neighbouring coefficients, respectively, with a number of coefficients per tuple being equal for tuples of the previous segments and tuples for the current segment, wherein the deriving the entropy coding context comprises computing the entropy coding context for the current coefficient by selecting a set of tuples of the previous segment in a manner so that a number of coefficients in the set of tuples in case of the number of coefficients of the previous segment and the number of coefficients of the current segment being different, is equal to the number of coefficients in the set of coefficients in case of the number of coefficients of the previous segment and the number of coefficients of the current segment not being different, and at least some of the set of tuples are selected out of the tuples of the previous segment and selected in a manner so that a spectral spacing between coefficients of the set of coefficients selected out of the coefficients of the previous segment is larger in case of the number of coefficients of the previous segment being larger than the number of coefficients of the current segment than in case of the number of coefficients of the previous segment being lower than the number of coefficients of the current segment, and computing the entropy coding context for the current coefficient on the basis of the number of coefficients of each of the set of tuples, and the entropy decoding comprises entropy decoding the current coefficient by entropy decoding a tuple which comprises the current coefficient using the entropy coding context.
This invention relates to video or image compression, specifically improving entropy coding efficiency in transform-based coding systems. The problem addressed is the mismatch in spectral coefficient distribution between adjacent segments (e.g., blocks or tiles) when their sizes differ, which can degrade compression performance. The solution involves spectrally subdividing both a previous segment and a current segment into tuples of neighboring coefficients, ensuring each tuple has the same number of coefficients across segments. When segment sizes differ, the method selects a subset of tuples from the previous segment such that the total number of coefficients in the subset matches the current segment's coefficient count. The selection prioritizes tuples with larger spectral spacing if the previous segment has more coefficients than the current one, and vice versa. The entropy coding context for a current coefficient is then derived from the number of coefficients in the selected tuples, and entropy decoding uses this context to decode the current coefficient's tuple. This approach ensures consistent context modeling across segments of varying sizes, improving compression efficiency.
32. The method for decoding an encoded audio stream of claim 31 , wherein the number of coefficients in the set of coefficients is four.
The invention relates to audio signal processing, specifically methods for decoding an encoded audio stream. The problem addressed is improving the efficiency and accuracy of audio decoding by optimizing the number of coefficients used in the decoding process. In audio encoding, signals are often transformed into a set of coefficients that represent the frequency components of the audio. During decoding, these coefficients are used to reconstruct the original audio signal. The invention specifies that the set of coefficients used in the decoding process should consist of exactly four coefficients. This precise number is chosen to balance computational efficiency with audio quality, ensuring that the decoding process is both fast and accurate. The method involves extracting these four coefficients from the encoded audio stream and applying an inverse transformation to convert them back into a time-domain audio signal. By limiting the number of coefficients to four, the method reduces the computational overhead while maintaining sufficient detail to reconstruct the audio with high fidelity. This approach is particularly useful in applications where processing power is limited, such as mobile devices or real-time audio processing systems. The invention ensures that the decoding process is optimized for both performance and audio quality.
Unknown
June 16, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.